Introduction

The Trust for Public Land (TPL) [1] tracks green space availability across U.S. using the proprietary ParkScore index. This index measures how well cities are meeting their residents’ need for public parks based on several categories:

Each city is awarded up to 100 points based on each one of these categories, and total of the scores is normalized to form an overall ParkScore rating. The data is taken from the Tidy Tuesday project [2]. In this report, we would like conduct exploratory data analysis to explore yearly trends across different categories of ParkScore and how the TPL has changed their methodology. We also explore importance of certain park features to their rankings. Finally, we analyze the newest 2021 data published by TPL to uncover how addition of the equity category to the ParkScore has affected cities’ rankings.

The map below shows ParkScore rankings of 95 cities in contiguous USA in 2020. A regional pattern is present - West Coast and Midwestern cities rank overall much higher (1-49 positions, in green) than Southwestern and Southern cities (50+). Minneapolis is the number one city that year and has overall been leading the rankings since 2012, along with .

Data cleaning and imputation

The raw data has already been processed by the TPL to form ParkScore, so there was little need to conduct etensive data cleaning. We have converted year variable as a factor, removed % and $ signs from 3 variable columns and converted them to numeric values, as well as renamed the variables for better plotting. More than 30% of the data was missing in 6 out of 28 variables: restroom, splashground, and park benches variables and their normalized scores. We omit them from the analysis and impute 9 other variables with 12-22% missingness. Assuming missing at random mechanism, they were imputed using mice package with Predictive Mean Matching (PMM) method. Diagnostic plots for imputation procedure showed good approximate convergence.

Data analysis

Exploratory data analysis

We investigated the overall variable distribution and the variable distribution over the years from 2012 to 2019. For variable distributions, most variables have an approximately normal distribution with some skewness towards right. Some potential outlines were observed. In the variable distribution over years, we observed a general increasing trend in scores for all attributes. For example, the park size score has an obvious larger number for year 2019 and 2020 compared with previous year. We will explain the reason for the increase in scores in next section.

Park Scores vs. Features

Top ranking cities across years

This figure shows the rank change of the top 12 parks from year 2015 to year 2020. Most cities’ ranks remained unchanged from year 2015 to year 2020. San Diego had a major drop in park ranking among these cities. When we further look into different score changes for San Diego, the scores for amenities decreased from 2015 to 2020, which is a major reason that caused the rank change.

Exploring equity in 2021

Addition of the equity score in 2021 has boosted rankings of some cities, such as Chicago, which moved from 10th position to 5th position. It also decreased ranking of Portland from 5-6th position to 10th position. To verify that this is due to the improved TPL methodology and inclusion of equity scores, we can calculate change across other variables from 2020 to 2021. As table below shows, this estimated change has been overall close to 0 for Chicago: the city lost a few points in park acreage and investment, but gained in park access and amenities. So equity score was likely the sole contributor to the improved ranking of Chicago. On the other hand, Portland has gained a lot of park acreage, but a decrease in park access, amentities, as well as a very average (=54) equity score did not help the city to maintain its park ranking.

Difference in scores between 2020 and 2021
Rank 2019 2020 2021 Equity in 2021
1 Washington, D.C. Minneapolis Washington 88
2 St. Paul Washington, D.C. St. Paul 72
3 Minneapolis St. Paul Minneapolis 60
4 Arlington, Virginia Arlington, Virginia Arlington 71
5 Portland Cincinnati Chicago 90
6 Irvine Portland San Francisco 63
7 San Francisco Irvine Irvine 64
8 Cincinnati San Francisco Cincinnati 68
9 New York Boston Seattle 77
10 Chicago Chicago Portland 54
Med_park_size Park_pct_city Pct_near_park Spend_per_resident Amenities
Chicago -6 1 13 -6 7.0
Portland 33 71 -12 0 -7.2

References

[1] https://www.tpl.org/parkscore

[2] https://github.com/rfordatascience/tidytuesday/blob/master/data/2021/2021-06-22/readme.md